Poor physician-patient communication is associated with problems such as decreased physical and mental health status, treatment compliance, diagnostic accuracy, and patient satisfaction. Yet efforts to encourage adoption of promising communication practices have been hindered by inconsistent definitions and a general lack of theory-based, causal explanations linking specific communication dynamics to medical visit outcomes. The proposed dissertation addresses this problem by mechanistically explaining why physicians'responses to patients'requests persistently vary, even under controlled experimental conditions. Because variation in how physicians respond to patients'requests is associated with the policy-relevant outcomes discussed above, explaining this variation gives insight into the broader relationship between communication and the outcomes of interest. We propose to explain variation in physicians'responses to patients'requests through role theory and welfare economics. Role theory implies that physicians engage in distinct patterns of recognizably physician-specific behaviors, some of which may match the physician Ideal Types influentially described by Emanuel and Emanuel (1992;E&E hereafter), and that variation across these patterns should explain variation in physicians'responses to patient requests. Welfare economics implies that spending time interacting with patients imposes real costs on physicians which increase as time elapses, so the total elapsed interaction time when a patient makes a request should also explain physician response variation. The proposed mechanisms naturally imply policy interventions. Physicians'communication behaviors are instilled during medical school and residency and are subject to modification, and framing time economically helps identify the incentives needed to achieve optimal time conditions within physician visits. Our analysis will focus on physicians'interactions with elderly patients, one of AHRQ's target populations, by utilizing preexisting data from the Assessment of Doctor-Elderly Patient Transactions (ADEPT) study, which includes video-recorded clinical encounters, patient and physician surveys, and additional clinical, demographic, and administrative data. Prior work suggests these data may contain 2000+ requests. We will determine the operation of roles by mapping the empirical probabilities of role-relevant behaviors onto latent traits using e.g. latent class analysis, and will derive the probabilities of roles conditioned on behaviors through Bayes'Theorem, evaluating whether the identified behavior patterns match those implied by E&E. We define request timing using elapsed time from the start of an interaction. We will code physicians'responses and patients'request types using the Taxonomy of Requests by Patients (TORP) algorithm, yielding a multivariate response variable and categorical request variables. We conclude by regressing physician response on any empirically identified roles, the E&E role types (if empirically valid), request timing, and interaction terms for all of them, controlling for request type, physician and patient affect, patient health status, reason for visit, and length of doctor-patient relationship. PUBLIC HEALTH RELEVANCE: Poor physician-patient communication is associated with poor medical care outcomes, yet efforts to improve such communication fall short. This project looks into this issue by explaining how physicians'responses to patients'requests vary in order to further AHRQ's mandate to improve healthcare safety/quality.